Google confirms that cosmic rays can cause uncorrectable errors in superconducting qubits
In June 2021, researchers at the University of Wisconsin-Madison speculated in a paper published in Nature that cosmic rays may be one of the causes of errors in superconducting qubits. Now, Google’s latest paper published in "Nature" shows that cosmic rays do cause superconducting qubits to go wrong.
In July of this year, when researchers tested error correction on Google's quantum processor , they noticed a strange phenomenon that the entire error correction scheme occasionally failed severely. They attributed this to background radiation, which is a combination of cosmic rays and the accidental decay of naturally occurring radioisotopes.
There were comments at the time that Google accidentally paid for an extremely expensive cosmic ray detector. But the researchers behind the processor take this issue very seriously and detail how radiation affects qubits in a new paper. They concluded that the problems caused by cosmic rays often occur and are enough to prevent error correction quantum computing work unless we find a way to limit the effects of the rays.
Google "Platanus" quantum processor
The influence of cosmic rays
Cosmic rays and radioactivity can also cause problems for classical computing hardware. This is because classical computers rely on moving and storing electric charges, and cosmic rays can induce electric charges when they hit matter.
In contrast, qubits store information in the form of the quantum state of an object—in the case of a Google processor, a superconducting coil connected to a resonator. Cosmic rays also affect them, but the mechanism is completely different.
The impact of cosmic rays also produces vibrational energy, which comes in the form of so-called phonons. These phonons can also be combined to form quasi-particles, where small groups of phonons come together and begin to behave like individual particles with different properties.
It is these quasiparticles that cause severe damage because they can exchange energy with quantum computing hardware. This includes the Cooper electron pair (another quasiparticle) that forms the basis of superconductivity, or the qubit itself, changing its state and breaking any entanglement.
If these phonons only affect a single qubit, then this situation will not be a problem-in fact, this is exactly what quantum error correction is about to deal with. Quantum error correction involves distributing quantum information among multiple entangled qubits, allowing the hardware to recognize when one of the qubits behaves abnormally.
The problem is that quasi-particles will not only affect the local area; on the contrary, they will be scattered around their origin and eventually affect multiple qubits. This should be enough to interfere with error correction. So, some people who implemented error correction in the early papers got together with some physicists to see if this really happened in quantum processing hardware.
Cosmic ray concept map
Error in the chip
To understand what happened, the Google team selected the 26 least error-prone qubits on its processor and set them all into a quantum state. The researchers can then leave the processor idle for a short period of time to see if the qubit is still in this state.
Cosmic ray impacts are easy to identify. After letting the processor idle for 100 microseconds, the typical background error rate is about 4 out of 26 qubits. When a cosmic ray happened to strike, about 24 qubits ended up in the wrong state—even though each qubit was about one millimeter away from its neighboring qubit.
To prove that this is due to quasiparticles, the researchers looked for a state dependence. It is expected that the quasiparticle will lose energy quickly, and therefore cannot transfer enough energy to elevate the qubit from the ground state to the excited state. But they can still absorb energy from the qubit, allowing the qubit in the excited state to fall back to the ground state.
Therefore, if quasiparticles are adjusting these interactions, when all qubits start in an excited state, you would expect more errors than when they all start in a ground state. This is exactly what the research team saw.
Because quantum processors can sample the state of qubits very quickly, the team can even track the propagation of errors in the processor. Initially, the error was mainly confined to the qubit closest to the cosmic ray impact. But even if the error rate here starts to decrease, qubits that are further away from the point of influence begin to see their error rate rise as the phonons spread across the chip. Before returning to normal, each qubit in the device usually sees its average error rate rise.
The problem we are facing
There is an obvious hint here. Error correction relies on configuring adjacent groups of qubits as a single logical unit, thereby protecting quantum information from any errors affecting a single qubit. But in the case of cosmic rays, all adjacent qubits are likely to make errors at the same time. In these situations, there is no realistic way to maintain error correction.
The key question then becomes whether these events happen frequently to affect calculations? If they are rare enough, we can discard those calculations that involve errors and start over. But this paper also studied this point, and the conclusion is not good. On average, the (quite small) quantum processor used here will have an error every 10 seconds. Most of the algorithms that we are interested in running on quantum computers can take several hours to complete. (This may seem long, but keep in mind that these calculations simply cannot be done on traditional hardware.)
Therefore, this is a problem for transmon superconducting qubits, and the paper shows that similar problems may affect other leading technologies. As we build larger processors to increase the number of qubits, the situation will only get worse. Is there anything I can do?
Unfortunately, the author of this paper had to turn to hypothesis here. They pointed out that astronomers encountered similar problems when designing imaging hardware and came up with ways to change the physical structure of the detector to limit the propagation of phonons. But it is not yet clear whether the technology used there is compatible with quantum processing hardware. But this paper seems to give people a good reason to find out.
link: https://www.nature.com/articles/s41586-021-03557-5
https://www.nature.com/articles/s41567-021-01432-8
https://www.nature.com/articles/s41586-021-03588-y
https://arstechnica.com/science/2021/12/cosmic-rays-can-swamp-error-correction-on-quantum-processors/
